ubd_kern.c

来自「h内核」· C语言 代码 · 共 1,656 行 · 第 1/3 页

	if(io_pid != -1) 
		os_kill_process(io_pid, 1);
}

__uml_exitcall(kill_io_thread);

static int ubd_file_size(struct ubd *dev, __u64 *size_out)
{
	char *file;

	file = dev->cow.file ? dev->cow.file : dev->file;
	return(os_file_size(file, size_out));
}

static void ubd_close(struct ubd *dev)
{
	if(ubd_do_mmap)
		physmem_forget_descriptor(dev->fd);
	os_close_file(dev->fd);
	if(dev->cow.file == NULL)
		return;

	if(ubd_do_mmap)
		physmem_forget_descriptor(dev->cow.fd);
	os_close_file(dev->cow.fd);
	vfree(dev->cow.bitmap);
	dev->cow.bitmap = NULL;
}

static int ubd_open_dev(struct ubd *dev)
{
	struct openflags flags;
	char **back_ptr;
	int err, create_cow, *create_ptr;

	dev->openflags = dev->boot_openflags;
	create_cow = 0;
	create_ptr = (dev->cow.file != NULL) ? &create_cow : NULL;
	back_ptr = dev->no_cow ? NULL : &dev->cow.file;
	dev->fd = open_ubd_file(dev->file, &dev->openflags, back_ptr,
				&dev->cow.bitmap_offset, &dev->cow.bitmap_len, 
				&dev->cow.data_offset, create_ptr);

	if((dev->fd == -ENOENT) && create_cow){
		dev->fd = create_cow_file(dev->file, dev->cow.file, 
					  dev->openflags, 1 << 9, PAGE_SIZE,
					  &dev->cow.bitmap_offset, 
					  &dev->cow.bitmap_len,
					  &dev->cow.data_offset);
		if(dev->fd >= 0){
			printk(KERN_INFO "Creating \"%s\" as COW file for "
			       "\"%s\"\n", dev->file, dev->cow.file);
		}
	}

	if(dev->fd < 0) return(dev->fd);

	if(dev->cow.file != NULL){
		err = -ENOMEM;
		dev->cow.bitmap = (void *) vmalloc(dev->cow.bitmap_len);
		if(dev->cow.bitmap == NULL){
			printk(KERN_ERR "Failed to vmalloc COW bitmap\n");
			goto error;
		}
		flush_tlb_kernel_vm();

		err = read_cow_bitmap(dev->fd, dev->cow.bitmap, 
				      dev->cow.bitmap_offset, 
				      dev->cow.bitmap_len);
		if(err < 0)
			goto error;

		flags = dev->openflags;
		flags.w = 0;
		err = open_ubd_file(dev->cow.file, &flags, NULL, NULL, NULL, 
				    NULL, NULL);
		if(err < 0) goto error;
		dev->cow.fd = err;
	}
	return(0);
 error:
	os_close_file(dev->fd);
	return(err);
}

static int ubd_new_disk(int major, u64 size, int unit,
			struct gendisk **disk_out)
			
{
	struct gendisk *disk;
	char from[sizeof("ubd/nnnnn\0")], to[sizeof("discnnnnn/disc\0")];
	int err;

	disk = alloc_disk(1 << UBD_SHIFT);
	if(disk == NULL)
		return(-ENOMEM);

	disk->major = major;
	disk->first_minor = unit << UBD_SHIFT;
	disk->fops = &ubd_blops;
	set_capacity(disk, size / 512);
	if(major == MAJOR_NR){
		sprintf(disk->disk_name, "ubd%c", 'a' + unit);
		sprintf(disk->devfs_name, "ubd/disc%d", unit);
		sprintf(from, "ubd/%d", unit);
		sprintf(to, "disc%d/disc", unit);
		err = devfs_mk_symlink(from, to);
		if(err)
			printk("ubd_new_disk failed to make link from %s to "
			       "%s, error = %d\n", from, to, err);
	}
	else {
		sprintf(disk->disk_name, "ubd_fake%d", unit);
		sprintf(disk->devfs_name, "ubd_fake/disc%d", unit);
	}

	/* sysfs register (not for ide fake devices) */
	if (major == MAJOR_NR) {
		ubd_dev[unit].pdev.id   = unit;
		ubd_dev[unit].pdev.name = DRIVER_NAME;
		platform_device_register(&ubd_dev[unit].pdev);
		disk->driverfs_dev = &ubd_dev[unit].pdev.dev;
	}

	disk->private_data = &ubd_dev[unit];
	disk->queue = ubd_queue;
	add_disk(disk);

	*disk_out = disk;
	return 0;
}

#define ROUND_BLOCK(n) ((n + ((1 << 9) - 1)) & (-1 << 9))

static int ubd_add(int n)
{
	struct ubd *dev = &ubd_dev[n];
	int err;

	if(dev->file == NULL)
		return(-ENODEV);

	if (ubd_open_dev(dev))
		return(-ENODEV);

	err = ubd_file_size(dev, &dev->size);
	if(err < 0)
		return(err);

	dev->size = ROUND_BLOCK(dev->size);

	err = ubd_new_disk(MAJOR_NR, dev->size, n, &ubd_gendisk[n]);
	if(err) 
		return(err);
 
	if(fake_major != MAJOR_NR)
		ubd_new_disk(fake_major, dev->size, n, 
			     &fake_gendisk[n]);

	/* perhaps this should also be under the "if (fake_major)" above */
	/* using the fake_disk->disk_name and also the fakehd_set name */
	if (fake_ide)
		make_ide_entries(ubd_gendisk[n]->disk_name);

	ubd_close(dev);
	return 0;
}

static int ubd_config(char *str)
{
	int n, err;

	str = uml_strdup(str);
	if(str == NULL){
		printk(KERN_ERR "ubd_config failed to strdup string\n");
		return(1);
	}
	err = ubd_setup_common(str, &n);
	if(err){
		kfree(str);
		return(-1);
	}
	if(n == -1) return(0);

 	spin_lock(&ubd_lock);
	err = ubd_add(n);
	if(err)
		ubd_dev[n].file = NULL;
 	spin_unlock(&ubd_lock);

	return(err);
}

static int ubd_get_config(char *name, char *str, int size, char **error_out)
{
	struct ubd *dev;
	char *end;
	int n, len = 0;

	n = simple_strtoul(name, &end, 0);
	if((*end != '\0') || (end == name)){
		*error_out = "ubd_get_config : didn't parse device number";
		return(-1);
	}

	if((n >= MAX_DEV) || (n < 0)){
		*error_out = "ubd_get_config : device number out of range";
		return(-1);
	}

	dev = &ubd_dev[n];
	spin_lock(&ubd_lock);

	if(dev->file == NULL){
		CONFIG_CHUNK(str, size, len, "", 1);
		goto out;
	}

	CONFIG_CHUNK(str, size, len, dev->file, 0);

	if(dev->cow.file != NULL){
		CONFIG_CHUNK(str, size, len, ",", 0);
		CONFIG_CHUNK(str, size, len, dev->cow.file, 1);
	}
	else CONFIG_CHUNK(str, size, len, "", 1);

 out:
	spin_unlock(&ubd_lock);
	return(len);
}

static int ubd_remove(char *str)
{
	struct ubd *dev;
	int n, err = -ENODEV;

	n = parse_unit(&str);

	if((n < 0) || (n >= MAX_DEV))
		return(err);

	dev = &ubd_dev[n];
	if(dev->count > 0)
		return(-EBUSY);	/* you cannot remove a open disk */

	err = 0;
 	spin_lock(&ubd_lock);

	if(ubd_gendisk[n] == NULL)
		goto out;

	del_gendisk(ubd_gendisk[n]);
	put_disk(ubd_gendisk[n]);
	ubd_gendisk[n] = NULL;

	if(fake_gendisk[n] != NULL){
		del_gendisk(fake_gendisk[n]);
		put_disk(fake_gendisk[n]);
		fake_gendisk[n] = NULL;
	}

	platform_device_unregister(&dev->pdev);
	*dev = ((struct ubd) DEFAULT_UBD);
	err = 0;
 out:
 	spin_unlock(&ubd_lock);
	return(err);
}

static struct mc_device ubd_mc = {
	.name		= "ubd",
	.config		= ubd_config,
 	.get_config	= ubd_get_config,
	.remove		= ubd_remove,
};

static int ubd_mc_init(void)
{
	mconsole_register_dev(&ubd_mc);
	return 0;
}

__initcall(ubd_mc_init);

static struct device_driver ubd_driver = {
	.name  = DRIVER_NAME,
	.bus   = &platform_bus_type,
};

int ubd_init(void)
{
        int i;

	devfs_mk_dir("ubd");
	if (register_blkdev(MAJOR_NR, "ubd"))
		return -1;

	ubd_queue = blk_init_queue(do_ubd_request, &ubd_io_lock);
	if (!ubd_queue) {
		unregister_blkdev(MAJOR_NR, "ubd");
		return -1;
	}
		
	if (fake_major != MAJOR_NR) {
		char name[sizeof("ubd_nnn\0")];

		snprintf(name, sizeof(name), "ubd_%d", fake_major);
		devfs_mk_dir(name);
		if (register_blkdev(fake_major, "ubd"))
			return -1;
	}
	driver_register(&ubd_driver);
	for (i = 0; i < MAX_DEV; i++) 
		ubd_add(i);
	return 0;
}

late_initcall(ubd_init);

int ubd_driver_init(void){
	unsigned long stack;
	int err;

	/* Set by CONFIG_BLK_DEV_UBD_SYNC or ubd=sync.*/
	if(global_openflags.s){
		printk(KERN_INFO "ubd: Synchronous mode\n");
		/* Letting ubd=sync be like using ubd#s= instead of ubd#= is
		 * enough. So use anyway the io thread. */
	}
	stack = alloc_stack(0, 0);
	io_pid = start_io_thread(stack + PAGE_SIZE - sizeof(void *), 
				 &thread_fd);
	if(io_pid < 0){
		printk(KERN_ERR 
		       "ubd : Failed to start I/O thread (errno = %d) - "
		       "falling back to synchronous I/O\n", -io_pid);
		io_pid = -1;
		return(0);
	}
	err = um_request_irq(UBD_IRQ, thread_fd, IRQ_READ, ubd_intr, 
			     SA_INTERRUPT, "ubd", ubd_dev);
	if(err != 0)
		printk(KERN_ERR "um_request_irq failed - errno = %d\n", -err);
	return(err);
}

device_initcall(ubd_driver_init);

static int ubd_open(struct inode *inode, struct file *filp)
{
	struct gendisk *disk = inode->i_bdev->bd_disk;
	struct ubd *dev = disk->private_data;
	int err = 0;

	if(dev->count == 0){
		err = ubd_open_dev(dev);
		if(err){
			printk(KERN_ERR "%s: Can't open \"%s\": errno = %d\n",
			       disk->disk_name, dev->file, -err);
			goto out;
		}
	}
	dev->count++;
	if((filp->f_mode & FMODE_WRITE) && !dev->openflags.w){
	        if(--dev->count == 0) ubd_close(dev);
	        err = -EROFS;
	}
 out:
	return(err);
}

static int ubd_release(struct inode * inode, struct file * file)
{
	struct gendisk *disk = inode->i_bdev->bd_disk;
	struct ubd *dev = disk->private_data;

	if(--dev->count == 0)
		ubd_close(dev);
	return(0);
}

static void cowify_bitmap(__u64 io_offset, int length, unsigned long *cow_mask,
			  __u64 *cow_offset, unsigned long *bitmap,
			  __u64 bitmap_offset, unsigned long *bitmap_words,
			  __u64 bitmap_len)
{
	__u64 sector = io_offset >> 9;
	int i, update_bitmap = 0;

	for(i = 0; i < length >> 9; i++){
		if(cow_mask != NULL)
			ubd_set_bit(i, (unsigned char *) cow_mask);
		if(ubd_test_bit(sector + i, (unsigned char *) bitmap))
			continue;

		update_bitmap = 1;
		ubd_set_bit(sector + i, (unsigned char *) bitmap);
	}

	if(!update_bitmap)
		return;

	*cow_offset = sector / (sizeof(unsigned long) * 8);

	/* This takes care of the case where we're exactly at the end of the
	 * device, and *cow_offset + 1 is off the end.  So, just back it up
	 * by one word.  Thanks to Lynn Kerby for the fix and James McMechan
	 * for the original diagnosis.
	 */
	if(*cow_offset == ((bitmap_len + sizeof(unsigned long) - 1) /
			   sizeof(unsigned long) - 1))
		(*cow_offset)--;

	bitmap_words[0] = bitmap[*cow_offset];
	bitmap_words[1] = bitmap[*cow_offset + 1];

	*cow_offset *= sizeof(unsigned long);
	*cow_offset += bitmap_offset;
}

static void cowify_req(struct io_thread_req *req, unsigned long *bitmap,
		       __u64 bitmap_offset, __u64 bitmap_len)
{
	__u64 sector = req->offset >> 9;
	int i;

	if(req->length > (sizeof(req->sector_mask) * 8) << 9)
		panic("Operation too long");

	if(req->op == UBD_READ) {
		for(i = 0; i < req->length >> 9; i++){
			if(ubd_test_bit(sector + i, (unsigned char *) bitmap))
				ubd_set_bit(i, (unsigned char *) 
					    &req->sector_mask);
                }
	}
	else cowify_bitmap(req->offset, req->length, &req->sector_mask,
			   &req->cow_offset, bitmap, bitmap_offset,
			   req->bitmap_words, bitmap_len);
}

static int mmap_fd(struct request *req, struct ubd *dev, __u64 offset)
{
	__u64 sector;
	unsigned char *bitmap;
	int bit, i;

	/* mmap must have been requested on the command line */
	if(!ubd_do_mmap)
		return(-1);

	/* The buffer must be page aligned */
	if(((unsigned long) req->buffer % UBD_MMAP_BLOCK_SIZE) != 0)
		return(-1);

	/* The request must be a page long */
	if((req->current_nr_sectors << 9) != PAGE_SIZE)
		return(-1);

	if(dev->cow.file == NULL)
		return(dev->fd);

	sector = offset >> 9;
	bitmap = (unsigned char *) dev->cow.bitmap;
	bit = ubd_test_bit(sector, bitmap);

	for(i = 1; i < req->current_nr_sectors; i++){
		if(ubd_test_bit(sector + i, bitmap) != bit)
			return(-1);
	}

	if(bit || (rq_data_dir(req) == WRITE))
		offset += dev->cow.data_offset;

	/* The data on disk must be page aligned */
	if((offset % UBD_MMAP_BLOCK_SIZE) != 0)
		return(-1);

	return(bit ? dev->fd : dev->cow.fd);
}

static int prepare_mmap_request(struct ubd *dev, int fd, __u64 offset,
				struct request *req,
				struct io_thread_req *io_req)
{
	int err;

	if(rq_data_dir(req) == WRITE){
		/* Writes are almost no-ops since the new data is already in the
		 * host page cache
		 */
		dev->map_writes++;
		if(dev->cow.file != NULL)
			cowify_bitmap(io_req->offset, io_req->length,
				      &io_req->sector_mask, &io_req->cow_offset,
				      dev->cow.bitmap, dev->cow.bitmap_offset,
				      io_req->bitmap_words,
				      dev->cow.bitmap_len);
	}
	else {
		int w;

		if((dev->cow.file != NULL) && (fd == dev->cow.fd))
			w = 0;
		else w = dev->openflags.w;

		if((dev->cow.file != NULL) && (fd == dev->fd))
			offset += dev->cow.data_offset;

		err = physmem_subst_mapping(req->buffer, fd, offset, w);
		if(err){
			printk("physmem_subst_mapping failed, err = %d\n",
			       -err);
			return(1);
		}
		dev->map_reads++;
	}
	io_req->op = UBD_MMAP;
	io_req->buffer = req->buffer;
	return(0);
}

/* Called with ubd_io_lock held */
static int prepare_request(struct request *req, struct io_thread_req *io_req)
{
	struct gendisk *disk = req->rq_disk;
	struct ubd *dev = disk->private_data;
	__u64 offset;
	int len, fd;

	if(req->rq_status == RQ_INACTIVE) return(1);

	if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
		printk("Write attempted on readonly ubd device %s\n", 
		       disk->disk_name);
		end_request(req, 0);
		return(1);
	}

	offset = ((__u64) req->sector) << 9;
	len = req->current_nr_sectors << 9;

	io_req->fds[0] = (dev->cow.file != NULL) ? dev->cow.fd : dev->fd;
	io_req->fds[1] = dev->fd;
	io_req->map_fd = -1;
	io_req->cow_offset = -1;
	io_req->offset = offset;
	io_req->length = len;
	io_req->error = 0;
	io_req->sector_mask = 0;

	fd = mmap_fd(req, dev, io_req->offset);